MemoryPilot
by Soflutionltd
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MEMORYPILOT_DB | No | Path to the SQLite database file. Defaults to ~/.MemoryPilot/memory.db; set it to point at a copy (e.g. for testing). | |
| MEMORYPILOT_MMR | No | MMR diversity over the finalists; `0` disables it. | on |
| MEMORYPILOT_TOOLS | No | Set to `full` to restore the full tool list instead of the lean 17 core tools (the default). | |
| MEMORYPILOT_VERIFY | No | Git-based memory verification; `off` turns it off. | on |
| MEMORYPILOT_SECRETS | No | Set to `reveal` to restore transparent decryption instead of returning secrets as masked pointers. | |
| FASTEMBED_CACHE_PATH | No | Overrides the model cache directory (default ~/.cache/fastembed/, Hugging Face layout; models are downloaded on first run). | |
| MEMORYPILOT_TEMPORAL | No | Query-time date window boost/demotion; `0` disables it (date entities are still extracted at ingest). | on |
| MEMORYPILOT_MMR_FLOOR | No | Cosine below which two candidates are not considered redundant by MMR. | 0.55 |
| SUPABASE_ACCESS_TOKEN | No | Supabase access token (scope database:read) used by the Supabase connector; otherwise read from the Keychain service `MemoryPilot`, account `supabase-access-token`. | |
| MEMORYPILOT_CONNECTORS | No | GitHub, deploy and Supabase production connectors; `off` turns them off. | on |
| MEMORYPILOT_MASTER_KEY | No | Master key wrapping the per-tenant vault keys (envelope encryption) when running the server binary. | |
| MEMORYPILOT_MMR_LAMBDA | No | MMR lambda: 1.0 is pure relevance, 0.0 pure diversity. | 0.4 |
| MEMORYPILOT_PUBLIC_URL | No | Public origin of the server, used for OAuth 2.1 with MCP clients (claude.ai, ChatGPT). | |
| MEMORYPILOT_CONSOLIDATE | No | Daily embedding-based consolidation of near-duplicate ephemeral memories; `off` disables it. | on |
| MEMORYPILOT_EMBED_MODEL | No | Embedding model: `jina` (jina-embeddings-v5-text-nano-retrieval int8, 768-dim) or `jina-q4`, or Apache-2.0 alternatives via fastembed (`e5-small`, `e5-base`, `e5-large`, `bge-m3`). Changing the model drops existing vectors and re-embeds in the background. | jina |
| MEMORYPILOT_TRUST_CROSS | No | Cross-encoder logit on the top hit above which abstention is suppressed. | 1.0 |
| MEMORYPILOT_CROSS_RERANK | No | `1`/`always` forces cross-encoder rerank on every query, `0`/`off` disables it; `adaptive` (default) fires on hard or non-English queries. | adaptive |
| MEMORYPILOT_HTTP_SECRETS | No | Set to `on` to allow `reveal_secret` over HTTP (refused by default). | |
| MEMORYPILOT_EMBED_THREADS | No | ONNX intra-op threads for the embedder. | min(4, cores) |
| MEMORYPILOT_ABSTAIN_COSINE | No | Cosine floor below which a search reports confidence.abstain = true (unless MEMORYPILOT_TRUST_CROSS is reached). | 0.35 |
| MEMORYPILOT_RERANKER_MODEL | No | Cross-encoder reranker model (e.g. `mmarco`, `jina-v2`, `gte-multilingual`, legacy fp32 fastembed models). | mmarco |
| MEMORYPILOT_RERANK_THREADS | No | ONNX intra-op threads for the cross-encoder reranker. | min(2, cores) |
| MEMORYPILOT_REVEAL_CONFIRM | No | Set to `none` to skip the macOS confirmation dialog when revealing a secret (audit log only). | |
| MEMORYPILOT_EMBED_POOL_SIZE | No | Number of embedding ONNX sessions in the pool (legacy fastembed models use 4, or 2 for 1024-dim). | 1 |
| MEMORYPILOT_MODEL_IDLE_SECS | No | Idle time after which the embedder and reranker sessions are dropped to free memory; `0` keeps them resident. | 600 |
| MEMORYPILOT_RERANK_POOL_SIZE | No | Number of cross-encoder ONNX sessions kept hot (each extra session costs ~120 MB). | 1 |
| MEMORYPILOT_CROSS_RERANK_TOP_K | No | Number of candidates the cross-encoder rescores. | 12 |
| MEMORYPILOT_READER_NULL_MARGIN | No | Extractive reader: the span logit must beat the no-answer logit by this much; negative values answer more often. | 0 |
| MEMORYPILOT_CROSS_RERANK_WEIGHT | No | Fusion weight given to the cross-encoder score against the RRF score. | 0.45 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Server capabilities have not been inspected yet.
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
No tools | |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
This server cannot be deployed
Maintenance
ActivityActive
ResponsivenessNo issues